Monday, August 24, 2009

The Mass Layoff Non-Seasonal Disconnect

The BLS considers a mass layoff event to be a condition where there are at least fifty initial claims for unemployment insurance originating from a single employer over a period of five consecutive weeks.

The recent (released on Friday) report showed a sizable "drop" on a seasonally adjusted basis. NPR's Matthew Katz' initial reaction was similar to mine:

The decline may just be because we've had so many mass layoffs already that we've exhausted them. Still, there's something about the nosedive that chart is taking that makes me happy. Let's hope it continues.
But looking at the data, it may be just another odd seasonal adjustment that was the cause of the goods news.

As can be seen above, the difference between the seasonal and non-seasonal adjusted figure was a whopping 130,000 jobs. According to the BLS:

Seasonal adjustment is the process of estimating and removing the effect on time series data of regularly recurring seasonal events such as changes in the weather, holidays, and the beginning and ending of the school year. The use of seasonal adjustment makes it easier to observe fundamental changes in time series, particularly those associated with general economic expansions and contractions.

Thus, most seasonally adjustments smooth out month over month changes, while year over year figures of seasonal and non-seasonal data tend to be closely aligned. Not so much the case here. The chart below shows year over year changes of both the seasonal and non-seasonal data. We see the large drop in seasonal data (the "good news" reported), but non-seasonal adjusted data shows year over year mass layoffs at a six-month high.

Source: BLS


  1. Good work yet again, and reminds me of the annualised q-o-q GDP data that most government like to report. Economists and governments alike are just using math to claim that "growth" and "end of recession". One should use yoy real GDP to see iof there is real growth, and to jedge whether the economy is better this year vs last, and not whether this quarter vs last. The amount of distorsion from seasonal adjustments and noise from one quarter to the next is just incredible. Reinforces this engineer's beliefthat most mathematically-weak students end up becoming economists and statisticians and politicians.

  2. Hey CEO - we resemble that remark.

    Jake - interesting and unusual. I quite worrying about NSA vs SA after tracking it on many series (GDP, PCE, Sales,Ind. Prod.,....) and finding out that they basically just overlaid one another for all practical purposes. That's the biggest aberration that I've ever seen. Thanks for calling it to your attention.

  3. Or it could be that many companies are now doing what IBM has been doing and laying off fewer at a time to stay under the radar and not report.